It's called an 'AI assistant': A strategic approach to the strengths and weaknesses of AI scribes

"Everything sparkles on the sales floor." You know the feeling. The newest thing is out, and for as long as it's new, the possibilities seem endless. New cars have "self-driving," new phones have "fitness tracking" and even your neighbor's fence is "getting fixed."

But every new thing comes with a reality. The self-driving function is limited to the interstate, the fitness-tracking phone still needs you to exercise and your neighbor's fence is ... a foregone conclusion.

For health care professionals, the sparkly new thing was the AI scribe. For the thousands of providers working long hours after clinic just to finish charts, a heavenly chorus was singing: "No more pajama time!" The program listens to a recording of the appointment, turns that recording into structured medical information such as SOAP notes and, voila, charting time shrinks. Reality was ready for this moment all the same.

AI scribes arrived on the scene promising to do the charting for you, but that wasn't quite the case. Was the note entered into the EHR, or did you have to copy and paste it in? Did the note pick up all the pertinent information? Did it hallucinate or enter inaccurate data? Does the provider have to spend as long reviewing the AI-generated note as it would have taken to write it?

Each of these questions scrapes the shine off health care technology's new thing. But this is not a lesson in pessimism; it is a moment to consider our perspective. It's called an AI assistant, not an AI manager, after all. No reasonable person would hire an office assistant and then expect that person to run the office.

From self-driving cars to the newest AI scribe, limitations merely define scope. Both work best when they are constrained to their proper use case. Knowing when to use a tool, and when not to, is the secret to efficiency in any professional environment. Planning for the boundaries of AI scribes now will help you avoid headaches later.

Identify your use case

Working with an AI scribe is like hiring an assistant: You need to match the job requirements to the new employee. If the employee can't meet the job requirements, that's their problem. If you ask them to complete a task they aren't qualified for, you're gambling. Identify what the tool does well, and keep it focused there.

Keeping good records is the name of the game. What specific roles are you filling? Are you targeting particular visit types? Document the targets for your new AI scribe and track its success rate. How often do the notes need edits, and where? How long does charting take compared with before the AI scribe? A list of measurable performance indicators makes it much easier to maximize effectiveness. Use the tool where it's working well, and scale it back where it falls short.

Eliminate overlapping workflows and non-uses

Redundancy is the death of any efficiency tool. If your AI scribe requires you to repeat its work, your workflow has gotten longer. These are non-uses, places where the AI scribe's capabilities don't match your needs and you would have been better off without it. The key is figuring out whether that section of the workflow can be eliminated. If your AI scribe vendor allows it, remove the portions of the workflow that aren't working. Cutting off the dead branches will define the appropriate boundaries of your AI scribe's effective use.

Don't just plug it in; reshape your workflow

Both the uses and non-uses depend on your approach to integration. If you try to force a new service into an old role without making adjustments, you may discover gaps, overlaps and errors along the margins. Plug-in workflows sound great, but they assume too much about your existing process. Lay the foundation for success by making sure the role the AI scribe fills actually matches its function.

If you're willing to revise your operations around the new tool's strengths and weaknesses, you'll waste far less time in the long run. The AI scribe operates only in its zone of efficiency and frees up time for you to fill in the remaining gaps, rather than trying to do everything and leaving you to redo everything anyway.

Take a moment to think of the AI scribe as an AI assistant. Assistant. Improve that assistant's chance of success by giving it an assistant's workload and holding it accountable to those expectations. When someone releases AI managers, we can reconsider that role.

Cooper Rock is an executive account manager at ScribeBrain, an ambient scribe company focused on merging revenue cycle management with efficient clinical documentation.